Follow-Up Mastery

The first answer is never the best answer. Not because the model is holding back — it genuinely cannot know what level of depth and specificity you need until you push back. Follow-up prompts are where research happens. The initial answer is just the starting point for a conversation that drills down to actual insight.

1. Challenge Assumptions "What assumptions are baked into this analysis? Which of them could be wrong?" This forces the model to surface its hidden premises. The most dangerous assumptions are the ones nobody questions.
2. Request Evidence "What specific evidence supports your top finding? Where is that evidence weakest?" This separates well-supported claims from plausible-sounding speculation.
3. Explore Contradictions "What is the strongest counterargument to your conclusion? Who would disagree and why?" This stress-tests the analysis. If the model cannot articulate a strong counterargument, the conclusion is either very strong or the model is not thinking hard enough.
4. Ask What You Are Missing "What important aspects of this topic did you not cover? What would a domain expert add that you left out?" This catches blind spots and expands the research surface area.

The "what am I missing" prompt is the most underused follow-up in AI-assisted research. It works because the model often has relevant knowledge that the original question did not activate. If you asked about market size, the model might not have mentioned regulatory risk — not because it does not know about it, but because you did not ask. "What important angles did we not explore?" opens doors you did not know existed.